[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Speech-25] speechd up to v0.29

I don't know if all of the changelog entries since 8/1 have been posted to
this list, so I'm posting those now.  From now on I'll only post the bits
that haven't been posted, and no more than once a day.  I'd appreciate
anybody taking the time to install the latest version & posting to the
list about how it goes.


Aug  1 14:41:29 EDT 1999 v0.15 Darxus - automatically loads "festival --server" &
                               when creating /dev/speech, makes it world writeable.
Aug  1 15:21:58 EDT 1999 v0.16 Darxus - more proper handling of pause
                               between executing "festival --server" &
                               when it starts accepting connections.
Aug  1 16:09:24 EDT 1999 v0.17 Darxus - corrected handling of absent
                               command line options.
Aug  1 19:11:45 EDT 1999 v0.18 Darxus - test for festival server before
                               attempting to start it.
Aug  1 16:13:06 PST 1999 v0.19 Michael - FIFO fixes, -q and -Q cmd line
                               options added, and command line help
Aug  1 19:37:37 EDT 1999 v0.20 Michael - open /dev/null readonly
Aug  1 19:42:06 EDT 1999 v0.21 Darxus - verify not only that $synth
                               isn't null, but actually valid.
Aug  1 17:02:17 PST 1999 v0.22 Michael - now daemonizes by default, and
                               cleaned up -Q
Aug  1 21:46:34 EDT 1999 v0.23 Darxus - read /etc/speechdrc & ~/.speechdrc.
Aug  1 23:25:36 EDT 1999 v0.24 Darxus - made speechdrc files work.  Oops.
Aug  3 20:40:16 EDT 1999 v0.25 Darxus - changed default festival function
                               to tts_file, added "or die" for write to
v0.26 6102 Aug  3 21:47:16 EDT 1999 
 1) Moved code to connect to festival out to connect_to_festival function.
 2) Allowed overriding of speechdrc file variables.

v0.27 6259 Aug  3 22:01:15 EDT 1999 
 Fixed "Broken pipe" bug discovered by Michael.
 Cought SIGPIPE when connection to Festival breaks & call

v0.28 6393 Aug  3 22:31:17 EDT 1999 
 1) Moved load_festival into connect_to_festival.
 2) Changed or's to if/else's in connect_to_festival.

v0.29 6247 Aug  3 22:51:15 EDT 1999 
 Eliminated redundant code in connect_to_festival.

I plan to reformat the whole CHANGELOG to be like the last 4 entries.

I've come accross an odd duplicate connection bug.

All connections to the festival server are now done by calling the
connect_to_festival function.  This is done 2 times -- once at the
beginning, and whenver a SIGPIPE signal is recieved (SIGPIPEs are
generated when the connection to the Festival server breaks).  

It looks like, when you have a connection to Festival, and "killall
festival", the 1st failed attempt to re-establish the connection (which
then attempts to re-start festival) causes another SIGPIPE signal, which
causes a second instance of the connect_to_festival function to be called.

There are, at that instant, 2 successful connections to the festival
server (according to the output from the server), then one of them
disconnects.  It doesn't seem to cause any problems, but doesn't look
right.  It all happens in less than 1 second.

Tried eliminating it by checking the contents of $handle, and failed.  And
I'm tired.

Michael isn't really happy about my solution to the Broken pipe bug
(waiting for the connection to get severed, then re-establishing it).
Tried to get him to discuss it here, but he's being stubborn :)
I don't want to misinterpret him (which is why I wanted him to post), but
I think he feels that it would be better to periodically send something
accross the pipe to keep festival from timing out the connection (if, in
fact, that's what's happening).  Basically, I think he wants to know
exactly why it's happening.  Which I think is nice, but I'm happy with my
solution (assuming it stands up to being left running overnight, which is
how the bug was replicated).  So I suggested that since he cares, he try
to fix it.  

I love opensource development :)

Nothing new has been posted to Freshmeat since I submitted speechd.  Guess
it has something to do w/ scoop (the Freshmeat guy) birthdaying today.

Anybody think it would be a bad idea to submit an update tomorrow ?  
Every day that there's an update available to be submitted ?  (but no more
than once a day)

PGP fingerprint = 03 5B 9B A0 16 33 91 2F  A5 77 BC EE 43 71 98 D4
            darxus@op.net / http://www.op.net/~darxus
                         Far Beyond Reason